Simulation experiment method and system for surface heat treatment of track link section

Abstract

The invention discloses a simulation experiment method and system for track link section surface heat treatment, and relates to the technical field of heat treatment simulation and optimizing.The simulation experiment method comprises the steps that basic data of a track link section heat treatment experiment are collected and preprocessed; constructing a coupling model of interaction of a thermal field, a stress field and a chemical field; training an intelligent learning model by using historical experimental data and a multi-physics field simulation result; in the simulation experiment process, the multi-physics field simulation data and the real-time collected data are combined, real-time error feedback parameters are obtained through the intelligent learning model, and heat treatment parameters are adjusted in real time according to the real-time error feedback parameters; and verifying the accuracy of a simulation result, and storing experimental data for repeated experiments and subsequent optimization. According to the method, heat treatment parameters can be dynamically adjusted, temperature distribution uniformity is improved, stress concentration is reduced, alloy component diffusion uniformity is improved, so that the fatigue life and mechanical performance of the track link section are remarkably improved, and high efficiency, reliability and industrial practicability are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of heat treatment simulation and optimization, and particularly to a simulation experiment method and system for the surface heat treatment of track links. Background Art

[0002] As a key load-bearing component of construction machinery and rail transit equipment, the surface performance of track links directly affects the wear resistance, fatigue life, and operation reliability of the equipment. Heat treatment technology is the main method to improve the surface performance of track links. By optimizing the heating, insulation, and cooling processes, it can have high hardness and anti-fatigue performance. In recent years, multi-physics field simulation technology has been widely used in the research of heat treatment processes. By constructing interaction models of the thermal field, stress field, and chemical field, the influence of different process parameters on the performance of track links is studied. At the same time, the rapid development of artificial intelligence technology provides new ideas for data-driven heat treatment process optimization. Combining historical experimental data and simulation results, using intelligent learning models to optimize heat treatment parameters in real time not only improves process stability but also reduces the experimental cycle and cost. However, existing research mostly focuses on the optimization of a single physical field and lacks comprehensive consideration of the interaction of multiple physical fields, which limits the simulation accuracy and the effect of parameter optimization.

[0003] Although the existing heat treatment process simulation technology has played an important role in optimizing the surface performance of track links, there are still many deficiencies in multi-physics field coupling modeling, intelligent optimization, and simulation accuracy. First, most traditional heat treatment simulation methods focus on a single physical field model of the thermal field and lack systematic consideration of the interaction relationship between the stress field and the chemical field, resulting in an incomplete comprehensive description of the heat treatment process. For example, the change of the stress field will significantly affect the diffusion behavior of chemical elements, and the change of the chemical field in turn affects the thermal field. Existing methods are difficult to accurately simulate these coupling relationships, affecting the accuracy of simulation results and the optimization effect of the process. Second, in terms of parameter optimization, existing technologies mostly rely on the offline analysis of experimental data and lack the dynamic optimization ability to combine simulation and real-time data feedback. This lag makes it difficult to quickly adjust heat treatment parameters according to process changes. In addition, due to the lack of support from intelligent learning models, the utilization efficiency of historical data and simulation results is low, and it is difficult to form a closed-loop optimization system. Finally, in terms of simulation verification and experimental data storage, existing technologies have not effectively integrated experimental data, simulation results, and optimization schemes, limiting the process repeatability and the efficiency of model improvement. Therefore, the existing technology cannot achieve accurate simulation, real-time optimization, and efficient verification of the surface heat treatment process of track links, and it is difficult to meet the requirements of engineering applications for high-performance track links. Summary of the Invention

[0004] In view of the above problems, the present invention is proposed.

[0005] Therefore, the technical problems solved by the present invention are as follows: existing simulation and optimization methods for track link heat treatment have problems such as insufficient construction of multi-physical field coupling models, lack of real-time data feedback and dynamic optimization capabilities, inability to efficiently integrate experimental data and simulation results, and how to achieve accurate simulation, parameter optimization, and efficient verification of the track link heat treatment process.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, an embodiment of the present invention provides a simulation experiment method for track link surface heat treatment, including collecting basic data of track link heat treatment experiments and performing preprocessing; Based on the collected and preprocessed data, constructing a coupling model of the interaction between the thermal field, stress field, and chemical field; Using historical experimental data and multi-physical field simulation results to train an intelligent learning model; During the simulation experiment process, combining multi-physical field simulation data and real-time collected data, obtaining real-time error feedback parameters through the intelligent learning model, and adjusting the heat treatment parameters in real time according to the real-time error feedback parameters; Verifying the accuracy of the simulation results and storing the experimental data for repeated experiments and subsequent optimization.

[0007] As a preferred solution of the simulation experiment method for track link surface heat treatment according to the present invention, wherein: the collection of basic data of track link heat treatment experiments includes material thermal physical property parameters, surface temperature data, mechanical property data, internal stress distribution data, chemical composition distribution data, heat treatment process parameters, fatigue life data, and multi-physical field dynamic interaction data.

[0008] As a preferred solution of the simulation experiment method for track link surface heat treatment according to the present invention, wherein: the preprocessing includes data denoising, missing data filling, outlier removal, data normalization, dynamic feature extraction, and data labeling.

[0009] As a preferred solution of the simulation experiment method for track link surface heat treatment according to the present invention, wherein: the construction of the coupling model of the interaction between the thermal field, stress field, and chemical field includes, based on the Fourier heat conduction equation, describing the spatio-temporal variation of the temperature distribution of track link materials during heat treatment, and optimizing the model by introducing the influence term of the chemical field on the thermal field to complete the construction of the thermal field model; Based on the stress balance equation, describing the material deformation and stress distribution caused by thermal stress during track link heat treatment, and optimizing the model by introducing the action term of the thermal gradient on the stress field to complete the construction of the stress field model; Based on the diffusion equation, describe the change in the concentration distribution of alloy elements during the heat treatment process of the track link, and optimize the model by introducing the coupling term of stress change to the diffusion process to complete the construction of the chemical field model; Couple the thermal field, stress field, and chemical field through the shared time dimension and spatial coordinates and define the boundary conditions and initial conditions; Use the finite element method to discretely solve the coupled model of the thermal field, stress field, and chemical field.

[0010] As a preferred scheme of the simulation experiment method for the surface heat treatment of the track link described in the present invention, wherein: the training intelligent learning model includes the collected historical experimental data and multi-physical field simulation results, including the spatio-temporal change data of the temperature distribution, stress distribution, and chemical composition concentration during the heat treatment process of the track link. Organize the collected historical experimental data and multi-physical field simulation results into a training set and a validation set, and the validation set is used to evaluate the generalization ability of the model; Extract features from the training data to obtain the feature data of the thermal field, stress field, and chemical field; Construct a deep neural network model containing multiple input layers and output layers, with the input being the feature data of the thermal field, stress field, and chemical field, and the output being the optimized heat treatment process parameters , and the loss function of the model is defined as: ; wherein, is the total number of time steps, is the th temperature distribution value in the simulation data, is the th stress value in the simulation data, is the th chemical concentration value in the simulation data, is the th temperature distribution value in the experimental data, the th stress value in the experimental data, is the th chemical concentration value in the experimental data; is the weight factor, dynamically adjusted to balance the contributions of different physical fields to the loss; Use the backpropagation algorithm to optimize the neural network weights; Adopt an adaptive learning rate optimization algorithm to adjust the learning rate to accelerate convergence; Dynamically adjust the loss function weight factor during the training process to adapt to the model training requirements at different stages; Calculate the prediction error of the model through the validation set; Adjust the number of layers, the number of neurons, and the activation function of the neural network according to the evaluation results; After the training is completed, the heat treatment process parameters output by the model The expression is: ; Where are the dynamic optimization factors of the thermal field, stress field, and chemical field.

[0011] As a preferred embodiment of the simulation experiment method for the surface heat treatment of the track link described in the present invention, wherein: the real-time adjustment of the heat treatment parameters includes, during the simulation experiment, real-time monitoring of the dynamic change data of the thermal field, stress field, and chemical field; Input the real-time collected data into the intelligent learning model, calculate the error feedback, and generate parameter adjustment suggestions. The formula is: ; Where is the real-time error feedback parameter; is the real-time collected temperature data, representing the actual temperature distribution of the track link during the heat treatment process; is the real-time collected stress data, representing the measured stress distribution of the track link during the heat treatment process; is the real-time collected chemical composition concentration data, representing the actual concentration distribution on the surface and inside of the track link; is the real-time simulated temperature, stress, and chemical concentration data; According to the error feedback result, optimize the heat treatment parameters in real time , and the adjustment formula is: ; Where is the optimized heat treatment parameter, is the current heat treatment parameter, is the learning rate, is the time step.

[0012] As a preferred embodiment of the simulation experiment method for the surface heat treatment of the track link described in the present invention, wherein: verifying the accuracy of the simulation results includes verifying the prediction accuracy of the simulation model by calculating the deviation between the real-time collected data and the simulation data. The specific formula is: ; Where is the comprehensive deviation at the th time step; When , it is a superior simulation, and the deviation is within an acceptable range, and there is no need to adjust the simulation model; When When it is in this range, it is a warning-level simulation. The deviation exceeds the expectation but is within the warning range, and local adjustment of the model parameters is required. When it is in this range, it is a non-conforming simulation. The deviation exceeds the warning range, and the simulation model needs to be reconstructed or the data preprocessing steps need to be adjusted. Among them, is the upper limit of the acceptable range of the comprehensive deviation . is the upper limit of the warning-level range of the comprehensive deviation . Define the overall accuracy of the simulation result as the root mean square error , and the calculation method is: . Among them, is the total number of time steps, which is used to evaluate the overall consistency between the simulation and the experiment. Define the coefficient of determination to measure the goodness of fit of the simulation model: . Among them, is the average value of the real-time temperature data.

[0013] In a second aspect, an embodiment of the present invention provides a simulation experiment system for the surface heat treatment of track links, including: Data acquisition and preprocessing module: Collect the basic data of the track link heat treatment experiment and perform preprocessing; Multi-physical field coupling model construction module: Based on the collected and preprocessed data, construct a coupling model of the interaction between the thermal field, stress field, and chemical field; Intelligent learning model training module: Use historical experimental data and multi-physical field simulation results to train the intelligent learning model; Real-time adjustment module: During the simulation experiment, combine multi-physical field simulation and intelligent learning model to adjust the heat treatment parameters in real time; Result verification and data storage module: Verify the accuracy of the simulation results and store the experimental data for repeated experiments and subsequent optimization.

[0014] Advantages of the present invention: By constructing a multi-physical field coupling model of the interaction between the thermal field, stress field, and chemical field, the present invention comprehensively considers the dynamic interaction relationship between different physical fields. For example, the influence of the change in the stress field on the diffusion of chemical elements, and the feedback effect of the chemical field on the heat conduction of the thermal field, thereby realizing a more accurate simulation of the track link heat treatment process and significantly improving the accuracy of the simulation results.

[0015] Combined with an intelligent learning model, historical experimental data and multi - physical field simulation results are used for model training, and during the heat treatment simulation process, the heat treatment parameters are adjusted in real - time through feedback. Compared with the traditional off - line analysis optimization method, the present invention can dynamically adjust process parameters such as heating rate, holding time, and cooling path, thereby improving the response speed and stability of the heat treatment process.

[0016] Through the combination of multi - physical field simulation and intelligent learning model, the present invention significantly reduces the dependence on actual experiments. The fusion of experimental data acquisition and simulation data forms a closed - loop optimization system, which can quickly screen out the optimal process parameters in the simulation stage, reducing the time and cost of a large number of repeated experiments.

[0017] The present invention realizes the deep integration of heat treatment technology, multi - physical field simulation technology and artificial intelligence technology, with innovation and practicality. The results can not only improve the manufacturing quality of track links, but also be extended and applied to the surface heat treatment optimization process of other mechanical parts, with high industrial application value. Brief Description of the Drawings

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings, where: Figure 1 It is the overall flowchart of a simulation experiment method for the surface heat treatment of a track link provided in the first embodiment of the present invention. Detailed Embodiments

[0019] To make the above - mentioned objects, features and advantages of the present invention more obvious and understandable, the detailed embodiments of the present invention will be described below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0020] Example 1, referring to Figure 1 , which is an embodiment of the present invention, provides a simulation experiment method for the surface heat treatment of a track link, including: S1: Collect the basic data of the track link heat treatment experiment and perform pre - processing.

[0021] In the heat treatment experiment of track links, data acquisition and preprocessing are crucial steps to ensure the accuracy of subsequent model construction and simulation. The basic data collected include material thermophysical properties, surface temperature data, mechanical property data, internal stress distribution data, chemical composition distribution data, process parameters during heat treatment, fatigue life data, and multi-physical field dynamic interaction data. The accuracy and integrity of these data directly affect the construction of multi-physical field coupling models and the training effect of intelligent learning models.

[0022] Furthermore, material thermophysical properties: including thermal conductivity, density, specific heat capacity, and latent heat of phase change of the material. These parameters are the basis for constructing the thermal field model and directly affect the accuracy of heat conduction calculation. For example, thermal conductivity determines the heat transfer efficiency of the material, and density and specific heat capacity affect the thermal inertia of the material. Accurately measuring these parameters can improve the accuracy of thermal field simulation; Surface temperature data: The data of the change of the surface temperature distribution of the track link over time measured by thermocouples or infrared thermal imagers. These data are the core indicators for verifying the accuracy of the heat treatment process, which are objective and easy to collect. Real-time monitoring of the surface temperature helps to control the heating and cooling rates and avoid deterioration of material properties caused by overheating or overcooling; Mechanical property data: The yield strength, elastic modulus, and ultimate tensile strength of the track link material measured by tensile and compression experiments. These data are closely related to the stress field model and are used to predict the mechanical properties of the track link after heat treatment. Accurate mechanical property data can improve the reliability of stress field simulation and ensure the safety of the track link under actual working conditions; Internal stress distribution data: The residual stress distribution of the track link during heat treatment measured by X-ray diffraction (XRD) or acoustic emission method. These data reflect the stress state after heat treatment and are an important verification basis for the accuracy of the stress field model. The existence of residual stress may affect the fatigue life of the track link, and accurate measurement helps to optimize the heat treatment process; Chemical composition distribution data: Wrap the XRF core component with a double-layer water cooling system (cooling water inlet temperature ≤ 20 °C, flow rate ≥ 5 L / min), combined with a silicon nitride ceramic protective layer (thickness 0.5 mm, temperature resistance ≥ 1500 °C) to ensure that the probe works stably when the surface temperature of the track link ≤ 1000 °C. The distance between the probe and the workpiece is fixed at 8 mm, supplemented by argon purging (flow rate 0.8 L / min) to remove the oxide layer and avoid interference from high-temperature surface contamination, and obtain the concentration distribution of alloy elements on the surface and inside of the track link. These data affect the accuracy of material properties and chemical field models. The change in the concentration of alloy elements will affect the hardness and wear resistance of the material, and accurate measurement can guide composition design and process adjustment; Process parameters of the heat treatment process: including heating rate, holding time, cooling rate, and the temperature and composition of the cooling medium. These parameters directly affect the heat treatment effect of the track link. Optimizing these parameters is the core task of heat treatment. Reasonable process parameters can ensure that the track link obtains an ideal organizational structure and performance; Fatigue life data: Based on fatigue experiments, obtain the fatigue life curve of the track link material under multiple loadings. Introduce fatigue life as an innovation point to expand the application scenarios of heat treatment simulation methods. Fatigue life data can be used to evaluate the durability of the track link under actual working conditions and guide product design and quality control; Multi-physical field dynamic interaction data: Record the dynamic change data of the thermal field, stress field, and chemical field during the heat treatment process through high-frequency sampling. Combine dynamic sampling and simulation analysis to reveal the real-time interaction of multi-physical fields. These data help to understand the coupling effect between different physical fields and optimize the accuracy of the model.

[0023] It should be noted that during the heat treatment process of the track link, the monitoring of the chemical composition distribution includes two levels: phased verification and real-time dynamic analysis. The present invention uses energy dispersive spectroscopy (EDS) and electron probe microanalysis (EPMA) as high-precision analysis means for chemical composition distribution data. Especially in the phased verification after heat treatment, it can accurately measure the concentration distribution of alloy elements on the surface and inside of the track link, providing important support for process parameter optimization and simulation model calibration. However, due to the insufficient real-time adaptability of traditional EDS and EPMA analysis equipment under dynamic high-temperature working conditions, in the present invention, on-line X-ray fluorescence spectroscopy (XRF) technology can also be used to realize the real-time monitoring of the surface chemical composition of the track link. On-line XRF has a response speed of seconds and non-contact analysis characteristics, and can meet the monitoring requirements under high-temperature dynamic environments.

[0024] Data preprocessing includes data denoising, missing data completion, outlier removal, data normalization, dynamic feature extraction, and data labeling; Data Denoising: Decompose and remove multi-scale noise through wavelet transform, which is applicable to high-frequency interference in material physical property parameters and real-time measurement data; in the track link heat treatment experiment, high-frequency noise may exist in the surface temperature data and stress data, such as environmental interference or equipment jitter. To improve the input accuracy of the simulation model, data denoising is required. The specific operation steps are as follows: 1. Input time series data (such as the curve of surface temperature changing with time) into the wavelet decomposition algorithm, and select the Daubechies wavelet (order 4, commonly used in physical experiment signal analysis); 2. Perform three-level decomposition to decompose the data into low-frequency components and multiple high-frequency components; 3. Use the soft threshold method to process the high-frequency components, and reconstruct the signal after eliminating the noise. The parameter selection is as follows: wavelet basis function: Daubechies wavelet (db4); decomposition level: 3; threshold: calculated based on the signal-to-noise ratio (SNR), set to 3 times the standard deviation of the high-frequency signal. There are high-frequency fluctuations of ±5°C in the original surface temperature data; after denoising, the fluctuation amplitude is reduced to ±0.5°C, and the signal-to-noise ratio is increased by 15 times; Missing Data Completion: Combine the hybrid strategy of interpolation method and deep learning model to ensure both the physical rationality of the completed data and enhance the prediction accuracy; missing values may occur in the internal stress distribution data of the track link due to sensor failure or insufficient sampling interval, which will affect the construction of the stress field model. The specific operation steps are as follows: 1. Perform linear interpolation on the missing data to initially fill in the small-range missing points in the continuous time series; 2. Use a deep learning model based on autoencoder to further optimize the interpolation result: construct the relationship between the input features (such as temperature distribution and time series) and the missing points; use historical experimental data to train the model to generate the completed result. The parameter selection is as follows: interpolation method: linear interpolation; number of autoencoder layers: 3 layers, fully connected neural network; activation function: ReLU; learning rate: 0.001. The proportion of missing stress data is 10%; after completion processing, the mean square error (MSE) of the completed data is reduced to 0.02, which is highly consistent with the real stress distribution; Outlier Removal: Double verification through the Z-score and density clustering algorithm improves the accuracy of outlier identification and ensures the statistical stability of the data; outliers may be generated in the fatigue life data and chemical composition concentration data due to equipment failure or human operation errors, which will lead to unstable model training results. The specific operation steps are as follows: 1. Calculate the mean and standard deviation of the data using the Z-score formula; 2. Use the density clustering algorithm (DBSCAN) to further detect isolated points, define the minimum number of points: 5; define the maximum neighborhood distance: 2 times the standard deviation of the data. In the fatigue life data, the original data contains extreme values (such as the life being 10 times the normal range); after removal, the data distribution conforms to the normal distribution, and the standard deviation decreases by 30%; Data normalization: Unify the dimension to reduce the model deviation caused by scale differences between different variables; the physical magnitudes of different data are different (e.g., the temperature unit is °C and the stress unit is MPa), and direct input may lead to unbalanced model weights. The specific operation steps are as follows: 1. Use the Min-Max normalization formula; 2. Calculate the normalization results for data such as temperature, stress, and chemical concentration respectively. For example, the original temperature range: [20°C, 1200°C]; after normalization: [0, 1]; the original stress range: [0 MPa, 500 MPa]; after normalization: [0, 1]; Dynamic feature extraction: Fourier transform is used to capture the frequency characteristics of dynamic interactions, and PCA optimizes the calculation performance through dimensionality reduction while retaining the main information; the multi-physical field interaction data contains a large amount of time series information, and the main features need to be extracted to reduce redundant data. The specific operation steps are as follows: 1. Perform Fourier transform on the time series data to extract the main frequency components (such as 0.1 Hz, which reflects the main dynamic behavior of the heat treatment heating rate); 2. Use principal component analysis (PCA) to reduce the dimensionality of the features: input dimension: 10 (including multiple variables such as temperature, stress, and chemical concentration); output dimension: 3 (total variance retention rate 95%). After dynamic feature extraction, the data storage volume is reduced by 70%, and the model training speed is increased by 30%; Data labeling: Classify data through process parameters, provide structured features for the input of the intelligent learning model, and improve the learning efficiency and accuracy of the model; the heat treatment process can be classified into multiple stages (such as heating, holding, cooling) for the model to learn. The specific operation steps are as follows: 1. Define stage labels according to heat treatment process parameters (such as time, temperature); 2. Attach the corresponding labels to each data to form a structured data set. After data labeling, the model prediction accuracy is increased by 20%.

[0025] S2: Based on the collected and preprocessed data, construct a coupled model of the interaction between the thermal field, stress field, and chemical field.

[0026] S21. Construct a thermal field model, a stress field model, and a chemical field model.

[0027] Construction of the thermal field model: Based on the Fourier heat conduction equation, describe the spatio-temporal variation of the temperature distribution of the track link material during heat treatment, and optimize the model by introducing the influence term of the chemical field on the thermal field. The specific expression is: ; where represents the temperature changing with time and is used to simulate the dynamic behavior of the thermal field; represents the concentration changing with time The rate of change is used to simulate the dynamic changes in the diffusion or reaction rate in the chemical field. Represents the concentration distribution of alloy elements in the track link material; Is the temperature distribution inside the track link material, which is a function of temperature with respect to time and space; Represents the time variable; Is the chemical coupling coefficient, used to quantify the feedback effect of the chemical field concentration change on the thermal field; Is the thermal diffusivity, defined as , Is the thermal conductivity, Is the material density, Is the specific heat capacity; Is the volumetric heat source term, representing the heat released due to chemical reactions or latent heat of phase change inside the material; Represents the Laplacian operator of the temperature field, describing the spatial change trend of heat conduction; Furthermore, by describing the temperature change through the Fourier heat conduction equation and introducing the coupling term of the chemical field, the feedback effect of alloy concentration change on the thermal field is quantified. The traditional thermal field model ignores the influence of heat release or absorption during chemical reactions on the temperature field, resulting in insufficient simulation accuracy of the temperature distribution; Even further, the thermal field model can accurately simulate the temperature distribution of the track link during the heat treatment process. Especially during the phase change and alloy element migration processes, the introduction of the chemical field compensates for the errors caused by ignoring the reaction heat effect in the traditional model and improves the prediction accuracy of the dynamic temperature changes.

[0028] Construction of the stress field model: Based on the stress equilibrium equation, it describes the material deformation and stress distribution caused by thermal stress during the heat treatment process of the track link, and optimizes the model by introducing the term of the effect of the thermal gradient on the stress field. The specific expression is: ; Among them, Is the stress tensor, describing the stress distribution generated in the track link material during the heat treatment process; Is the body force density, representing the distribution of external forces or internal force fields; Is the thermal gradient coupling coefficient, describing the degree of influence of temperature change on the stress field; Is the divergence of the stress tensor, used to calculate the stress change inside the material; Is the temperature gradient, used to describe the change in temperature distribution in space; Furthermore, the traditional stress field model does not fully consider the dynamic coupling relationship between the temperature field and the stress field, and it is difficult to reflect the thermal stress distribution in actual working conditions; Furthermore, the model can simulate the stress concentration phenomenon during rapid cooling or heating processes and accurately predict the residual stress distribution of the track link. The introduction of the thermal gradient improves the stress simulation ability of the model in locally supercooled or superheated regions, which helps optimize the heat treatment process and avoid cracks and deformations caused by stress concentration.

[0029] Construction of the chemical field model: Based on the diffusion equation, describe the change in the concentration distribution of alloy elements during the heat treatment process of the track link, and optimize the model by introducing a coupling term of stress change to the diffusion process. The specific expression is: ; where is the diffusion coefficient, representing the diffusion rate of alloy elements, which varies dynamically with temperature; is the Laplace operator of the concentration field, describing the spatial diffusion trend of the concentration distribution; is the reaction rate function, representing the influence of alloy concentration and temperature on chemical reactions; is the stress coupling coefficient, representing the influence of stress change on the diffusion behavior of alloy elements; is the rate of change of stress with time, representing the influence of the dynamic stress field on chemical diffusion; Furthermore, the stress coupling term improves the model's ability to describe the microstructure evolution process of the track link, such as the influence of stress gradient on the diffusion path. The optimized chemical field model provides data support for the composition design and heat treatment process improvement of high-performance track link materials.

[0030] S22. Couple the thermal field model, stress field model, and chemical field model through the shared time dimension and spatial coordinates and define the boundary conditions and initial conditions; The boundary conditions and initial conditions are defined as follows: Thermal field boundary conditions: Specify the convective heat transfer coefficient and ambient temperature of the track link surface to reasonably simulate the heat transfer process in the actual process; Stress field boundary conditions: Apply mechanical constraint boundary conditions to ensure the continuity of stress at the boundary, especially at the connection points or support areas; Chemical field boundary conditions: Set the diffusion boundary flux of alloy elements to simulate the migration behavior of elements on the surface and inside.

[0031] S23. Use the finite element method to discretize and solve the coupled model of the thermal field, stress field, and chemical field.

[0032] It should be noted that through unified time and space coupling, the model can truly reproduce the interaction of multiple physical fields during the heat treatment process of track links, improve the physical authenticity of the simulation, and the introduction of the finite element method significantly enhances the solution efficiency of the model under complex boundary conditions, providing support for rapid process optimization in industrial scenarios.

[0033] S3: Use historical experimental data and multi-physical field simulation results to train an intelligent learning model.

[0034] The collected historical experimental data and multi-physical field simulation results include the spatio-temporal variation data of temperature distribution, stress distribution, and chemical composition concentration during the heat treatment process of track links. Organize the collected historical experimental data and multi-physical field simulation results into a training set and a validation set, and the validation set is used to evaluate the generalization ability of the model.

[0035] Extract features from the training data to obtain feature data of the thermal field, stress field, and chemical field.

[0036] Construct a deep neural network model with multiple input layers and output layers. Its input is the feature data of the thermal field, stress field, and chemical field, and the output is the optimized heat treatment process parameters. , the loss function of the model is defined as: ; where is the total number of time steps, is the th temperature distribution value in the simulation data, is the th stress value in the simulation data, is the th chemical concentration value in the simulation data, is the th temperature distribution value in the experimental data, the th stress value in the experimental data, is the th chemical concentration value in the experimental data; is the weight factor, dynamically adjusted to balance the contributions of different physical fields to the loss.

[0037] Furthermore, to improve the comprehensiveness and accuracy of the intelligent learning model training, the present invention constructs a training set by combining historical experimental data and multi-physics field simulation data. The historical experimental data covers characteristic data such as temperature distribution, stress distribution, and chemical composition concentration of the track link under different heat treatment conditions. These data are from actual operations and have a high degree of credibility. The simulation data is obtained through multi-physics field model calculations, covering dynamic characteristics (such as the influence of thermal gradient on stress distribution) that are difficult to measure or collect in the experiment, and effectively complements the experimental data.

[0038] Use the backpropagation algorithm to optimize the neural network weights; adopt an adaptive learning rate optimization algorithm to adjust the learning rate to accelerate convergence; dynamically adjust the loss function weight factor during the training process , to meet the model training requirements at different stages.

[0039] Furthermore, in the present invention, the weight factors involved are consistent. This consistency ensures the unity of the optimization goal, that is, the relative importance of the thermal field, stress field, and chemical field remains coordinated during the overall optimization process. During the training stage of the intelligent learning model, the weight factors are normalized ( ), so that the optimization of the model has balance and interpretability; during the actual application stage, the weight factors are dynamically adjusted according to specific process goals, and the specific performance of different physical fields is preferentially optimized through reasonable allocation of weight values, such as fatigue life, temperature uniformity, or alloy composition distribution, etc. The present invention ensures the coherence of the optimization process by unifying the definition and dynamic adjustment mechanism of the weight factors in different calculation modules, and at the same time provides flexibility and pertinence for process optimization.

[0040] It should be noted that at the initial stage of training, a higher weight is given to the thermal field error ( ), because the temperature distribution is the most sensitive to the adjustment of process parameters. In the later optimization, gradually increase the weights of the stress field ( ) and the chemical field ( ) to ensure that the model can comprehensively optimize the accuracy of different physical fields.

[0041] Calculate the prediction error of the model through the validation set.

[0042] Adjust the number of layers, the number of neurons, and the activation function of the neural network according to the evaluation results.

[0043] After the training is completed, the heat treatment process parameters output by the model The expression is: ; Among them, are the dynamic optimization factors of the thermal field, stress field, and chemical field. The dynamic adjustment factor Ensuring the contribution ratio of different physical fields to parameter optimization at different time periods enhances the robustness of parameter adjustment.

[0044] S4: During the simulation experiment, combining multi - physical - field simulation data and real - time acquisition data, obtaining real - time error feedback parameters through an intelligent learning model, and adjusting the heat treatment parameters in real time according to the real - time error feedback parameters.

[0045] During the simulation experiment, dynamically monitor the data of the thermal field, stress field, and chemical field; Input the real - time acquisition data into the intelligent learning model, calculate the error feedback and generate parameter adjustment suggestions. The formula is: ; where, is the real - time error feedback parameter; is the real - time acquired temperature data, representing the actual temperature distribution of the track link during heat treatment; is the real - time acquired stress data, representing the measured stress distribution of the track link during heat treatment; is the real - time acquired chemical composition concentration data, representing the actual concentration distribution on the surface and inside of the track link; is the real - time simulation - calculated temperature, stress, and chemical concentration data.

[0046] It should be noted that the optimization of traditional heat treatment process parameters usually depends on the fixed values set before the experiment and cannot be dynamically adjusted according to real - time data. Existing real - time optimization technologies mostly focus on a single physical field and ignore the coupling relationship between multi - physical fields. Through the function to dynamically quantify the multi - physical - field error, it can sensitively capture the deviation changes during the heat treatment process and provide a basis for real - time optimization. The introduction of the weight factor enables the error feedback to be flexibly adjusted according to actual needs, improving the accuracy and adaptability of parameter adjustment.

[0047] According to the error feedback result, optimize the heat treatment parameters in real time , and the adjustment formula is: ; where, is the optimized heat treatment parameter, is the current heat treatment parameter, is the learning rate, is the time step.

[0048] It should be noted that traditional heat treatment parameter optimization methods are often based on the results of offline analysis and are difficult to adapt to the dynamic changes in complex process environments. and The introduction enables the parameter adjustment process to have convergence and stability, and can respond to real-time changes while ensuring process stability. The dynamic adjustment formula can significantly reduce defects (such as overheating, stress concentration, non-uniform chemical composition) during the heat treatment process and improve the performance consistency of track links.

[0049] Application scenarios of the real-time adjustment function include that when it is monitored that the temperature distribution of the track link at a certain moment (such as ) shows overheating, the real-time adjustment formula can avoid the coarsening of the microstructure caused by overheating by reducing the heating rate of the relevant parameters. For the real-time deviation feedback of the chemical composition concentration (such as ), the model can ensure the uniform distribution of alloying elements by optimizing the cooling rate and medium composition. The real-time monitoring of the stress field deviation (such as ) can dynamically adjust the holding time or cooling path when detecting the trend of residual stress concentration to prevent crack generation.

[0050] The design reflects the coupled feedback of multi-physical field errors in the time dimension, captures the dynamic changes of errors through partial derivative calculations, and realizes more precise process control. The heat treatment parameter adjustment formula can effectively combine real-time data with model simulation results to form a closed-loop control system, providing an efficient and stable intelligent optimization scheme for industrial heat treatment processes.

[0051] S5: Verify the accuracy of the simulation results and store the experimental data for repeated experiments and subsequent optimization.

[0052] Verify the prediction accuracy of the simulation model by calculating the deviation between the real-time collected data and the simulation data. The specific formula is: ; where is the comprehensive deviation at the th time step.

[0053] When , it is a superior simulation, the deviation is within the acceptable range, and there is no need to adjust the simulation model; when , it is a warning-level simulation, the deviation exceeds the expectation but is within the warning range, and local model parameters need to be adjusted; when , it is an unqualified simulation, the deviation exceeds the warning range, and the simulation model needs to be reconstructed or the data preprocessing steps need to be adjusted.

[0054] where is the upper limit of the acceptable range of the comprehensive deviation , that is, within this range, the consistency between the simulation results and the experimental data is considered to be good enough, and there is no need to further adjust the simulation model or process parameters; is the comprehensive deviation The upper limit of the warning level range, that is, when exceeding this range, the deviation between the simulation results and the experimental data is too large to directly guide the actual process, and it is necessary to readjust the simulation model or preprocess the experimental data.

[0055] Define the overall accuracy of the simulation results as the root mean square error , and the calculation method is: ; Among them, is used to evaluate the overall consistency between the simulation and the experiment.

[0056] When the value is small, it indicates that the model and the experiment have good consistency in the global range.

[0057] It can be used to evaluate the influence of different parameter combinations on the simulation accuracy and provide a basis for subsequent process optimization.

[0058] As a global index, the root mean square error makes up for the limitations of a single time step and provides an evaluation tool for the global performance of the model.

[0059] Define the coefficient of determination to measure the goodness of fit of the simulation model: ; Among them, is the average value of the real-time temperature data.

[0060] By comparing the sum of squares of the simulation errors and the sum of squares of the deviations of the experimental values, the relative fitting ability of the model is defined, the closer it is to 1, the better the fitting effect.

[0061] It should be noted that the present invention stores the experimental data and the simulation data, which helps to realize subsequent repeated experiments and model improvement. The stored data can be used to cross-batch verify the stability of the simulation model and improve the generality of process optimization.

[0062] Example 2, which is the second example of the present invention, is different from the previous example in that: When the above-mentioned function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the current technical solution can be embodied in the form of a software product. The current computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0063] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

[0064] More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection portion (electronic device) having one or more wirings, a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or otherwise processing as appropriate, and then storing it in a computer memory.

[0065] Embodiment 3, an embodiment of the present invention, provides a simulation experiment system for the surface heat treatment of a track link, including a data acquisition and preprocessing module, a multi-physical field coupling model construction module, an intelligent learning model training module, a real-time adjustment module, and a result verification and data storage module.

[0066] Data acquisition and preprocessing module: Collect the basic data of the track link heat treatment experiment and perform preprocessing.

[0067] Multi-physical field coupling model construction module: Based on the collected and preprocessed data, construct a coupling model of the interaction between the thermal field, stress field, and chemical field.

[0068] Intelligent learning model training module: Use historical experimental data and multi-physical field simulation results to train the intelligent learning model.

[0069] Real-time adjustment module: During the simulation experiment, combine the multi-physical field simulation data and real-time collected data, obtain the real-time error feedback parameters through the intelligent learning model, and adjust the heat treatment parameters in real time according to the real-time error feedback parameters.

[0070] Result verification and data storage module: Verify the accuracy of the simulation results and store the experimental data for repeated experiments and subsequent optimization.

[0071] Example 4, an embodiment of the present invention, provides a simulation experiment method for the surface heat treatment of track links. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation / contrast experiments.

[0072] In order to verify the effect of the simulation experiment method for the surface heat treatment of track links of the present invention, experimental simulation and data analysis of track link heat treatment were carried out. The experimental object was a track link specimen with standard dimensions, made of high-strength alloy steel, and the size specification was 50mm×20mm×10mm. The equipment required for the experiment included high-precision thermocouples, infrared thermal imagers, multi-axis mechanical testers, X-ray diffractometers (XRD), energy dispersive spectrometers (EDS), and high-performance computing clusters.

[0073] The experiment was divided into the following five steps: First, conduct a heat treatment experiment on the track link specimen, and collect its thermal physical properties parameters (such as thermal conductivity, specific heat capacity, density), surface temperature distribution, internal stress distribution, and chemical composition concentration data. Use thermocouples and infrared thermal imagers to monitor the temperature changes of the specimen during the heating and cooling processes, obtain the stress distribution data using XRD, and measure the chemical composition concentration distribution on the surface and inside through EDS. The experimental data was processed by denoising, missing value filling, and dynamic normalization to form a standardized data set, which was used as the input of the multi-physical field model.

[0074] Based on the collected data, a multi-physical field model coupling the thermal field, stress field, and chemical field was constructed. The model shared the time dimension and spatial coordinates and was discretely solved using the finite element method. To improve the calculation efficiency, the boundary conditions and initial conditions were defined in detail.

[0075] Utilize the multi - physical field simulation results and historical experimental data to train a deep neural network model. In the feature extraction stage, the dynamic time warping algorithm is used to uniformly standardize the temporal features of the thermal field, stress field, and chemical field. During the model training process, the weight factors are dynamically adjusted to optimize the heat treatment parameters.

[0076] During the simulation experiment, combine the multi - physical field simulation results with the heat treatment data collected in real - time, and use the error feedback mechanism to adjust the heat treatment parameters. According to the deviation level division strategy, locally optimize the areas with excessive temperature, stress concentration, or abnormal chemical composition distribution, and adjust the heating rate and cooling path in real - time.

[0077] After the experiment, compare the real - time collected data with the simulation data, and calculate the root mean square error (RMSE) and the coefficient of determination ( ), to verify the accuracy of the simulation results. All data are stored in the database for subsequent experiments and process optimization.

[0078] The experimental reference data are shown in Table 1.

[0079] Table 1 Experimental reference data

[0080] It can be observed from the tabular data that: the surface temperature of the specimens is controlled within the range of 815 - 823 °C, with small fluctuations, indicating good uniformity of temperature distribution during the heat treatment process. This benefits from the accurate calculation of heat conduction and chemical reaction heat by the multi - physical field model.

[0081] The internal stress distribution of the specimens is controlled within the range of 148 - 153 MPa, and the stress concentration phenomenon can be reduced by more than 20% compared with the traditional method. The multi - physical field model and the real - time parameter adjustment mechanism effectively reduce the risk of stress overload.

[0082] The chemical concentration fluctuations of the specimens are small, all remaining within 0.84 - 0.89 wt%, proving the superiority of the present invention in the control of alloy element diffusion.

[0083] The fatigue life of the specimens generally exceeds 495000 cycles, and the best reaches 520000 cycles.

[0084] Verified by this embodiment, the heat treatment simulation method of the track link of the present invention shows significant advantages in terms of temperature uniformity, stress optimization, and fatigue life improvement. The experimental data and analysis further prove that the present invention effectively improves the deficiencies of the existing technology and has innovation and practicality.

[0085] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A simulation experimental method for heat treatment of the surface of a track link, characterized in that: include: Collect basic data of heat treatment experiment of track link and perform preprocessing; Based on the collected and preprocessed data, a coupled model of the interaction between thermal field, stress field and chemical field is constructed; Use historical experimental data and multi-physics field simulation results to train intelligent learning models; In the simulation experiment, the multi-physics field simulation data and real-time acquisition data are combined to obtain real-time error feedback parameters through the intelligent learning model, and the heat treatment parameters are adjusted in real time according to the real-time error feedback parameters; Verify the accuracy of simulation results and store experimental data for repeated experiments and subsequent optimization.

2. The simulation experimental method for heat treatment of the surface of the track link according to claim 1, characterized in that: The basic data collected from the chain track segment heat treatment experiment include material thermophysical property parameters, surface temperature data, mechanical property data, internal stress distribution data, chemical composition distribution data, heat treatment process parameters, fatigue life data and multi-physical field dynamic interaction data.

3. The simulation experimental method for heat treatment of the surface of the track link according to claim 2, characterized in that: The preprocessing includes data denoising, missing data completion, outlier removal, data normalization, dynamic feature extraction and data labeling.

4. The simulation experimental method for heat treatment of the surface of the track link according to claim 3, characterized in that: The construction of the coupling model of the interaction between the thermal field, the stress field and the chemical field includes describing the temporal and spatial changes of the temperature distribution of the track link material during the heat treatment process based on the Fourier heat conduction equation, and completing the construction of the thermal field model by introducing the influence term of the chemical field on the thermal field to optimize the model; Based on the stress balance equation, the material deformation and stress distribution caused by thermal stress during the heat treatment of the track link are described, and the stress field model is constructed by introducing the effect of thermal gradient on the stress field to optimize the model. Based on the diffusion equation, the concentration distribution change of alloy elements during the heat treatment of the track link is described, and the construction of the chemical field model is completed by introducing the coupling term optimization model of the stress change to the diffusion process; The thermal, stress and chemical fields are combined through a shared time dimension and spatial coordinates Carry out coupling and define boundary conditions and initial conditions; The finite element method is used to discretize and solve the coupled model of thermal field, stress field and chemical field.

5. The simulation experimental method for heat treatment of the surface of the track link according to claim 4, characterized in that: The training intelligent learning model includes collecting historical experimental data and multi-physics field simulation results including the temperature distribution, stress distribution and spatiotemporal variation data of chemical composition concentration during the heat treatment of the track link, organizing the collected historical experimental data and multi-physics field simulation results into a training set and a validation set, and the validation set is used to evaluate the generalization ability of the model; Extract features from the training data to obtain feature data of thermal field, stress field and chemical field; Construct a deep neural network model with multiple input layers and output layers. Its input is the characteristic data of thermal field, stress field and chemical field, and its output is the optimized heat treatment process parameters. , the loss function of the model Defined as: ； in, is the total number of time steps, The simulation data The temperature distribution value, The simulation data stress value, The simulation data Chemical concentration values, The experimental data The temperature distribution value, The experimental data stress value, The experimental data Chemical concentration value; is a weighting factor that is dynamically adjusted to balance the contribution of different physical fields to the loss; Optimize neural network weights using back-propagation algorithm; Adopt adaptive learning rate optimization algorithm to adjust the learning rate to accelerate convergence; Dynamically adjust the loss function weight factor during training , to meet the model training needs at different stages; Calculate the prediction error of the model using the validation set; Adjust the number of layers, number of neurons and activation function of the neural network according to the evaluation results; After training is completed, the model outputs the heat treatment process parameters The expression is: ； in, Dynamic optimization factors for thermal, stress and chemical fields.

6. The simulation experimental method for heat treatment of the surface of the track link according to claim 5, characterized in that: The real-time adjustment of the heat treatment parameters includes, during the simulation experiment, real-time monitoring of dynamic change data of the thermal field, stress field and chemical field; The real-time collected data is input into the intelligent learning model, the error feedback is calculated and the parameter adjustment suggestions are generated. The formula is: ； in, is the real-time error feedback parameter; It is the temperature data collected in real time, indicating the actual temperature distribution of the track link during the heat treatment process; It is the stress data collected in real time, indicating the stress distribution of the track link measured during the heat treatment process; It is the chemical composition concentration data collected in real time, indicating the actual concentration distribution on the surface and inside of the track segment; Temperature, stress, and chemical concentration data calculated for real-time simulation; Optimize heat treatment parameters in real time based on error feedback results , the adjustment formula is: ； in, For the optimized heat treatment parameters, is the current heat treatment parameter, is the learning rate, is the time step.

7. The simulation experimental method for heat treatment of the surface of the track link according to claim 6, characterized in that: The verification of the accuracy of the simulation results includes verifying the prediction accuracy of the simulation model by calculating the deviation between the real-time collected data and the simulation data. The specific formula is: ； in, For the The combined deviation of the time steps; when When , it is an excellent simulation, the deviation is within the acceptable range, and there is no need to adjust the simulation model; when When , it is a warning level simulation, the deviation exceeds the expectation but is within the warning range, and the model parameters need to be adjusted locally; when When , it is an unqualified simulation, the deviation exceeds the warning range, and the simulation model needs to be reconstructed or the data preprocessing steps need to be adjusted; in, The comprehensive deviation The upper limit of the acceptable range; The comprehensive deviation The upper limit of the warning level range; The overall accuracy of the simulation results is defined as the root mean square error , calculated as: ； in, Used to evaluate the overall consistency between simulation and experiment; Defining the coefficient of determination Measure the goodness of fit of the simulation model: ； in, It is the average value of real-time temperature data.

8. A simulation experiment system for heat treatment of the surface of a track segment, used for implementing the simulation experiment method for heat treatment of the surface of a track segment as claimed in any one of claims 1 to 7, characterized in that: include: Data acquisition and preprocessing module: collects basic data of track link heat treatment experiment and performs preprocessing; Multi-physics field coupling model building module: Based on the collected and pre-processed data, a coupling model of the interaction between thermal field, stress field and chemical field is built; Intelligent learning model training module: Use historical experimental data and multi-physics field simulation results to train intelligent learning models; Real-time adjustment module: In the simulation experiment, the multi-physics field simulation data and real-time acquisition data are combined to obtain real-time error feedback parameters through the intelligent learning model, and the heat treatment parameters are adjusted in real time according to the real-time error feedback parameters; Result verification and data storage module: Verify the accuracy of simulation results and store experimental data for repeated experiments and subsequent optimization.

Images